[unix-history] / usr / src / sys / ufs / ffs / ufs_disksubr.c

/*
 * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
 * All rights reserved.
 *
 * %sccs.include.redist.c%
 *
 *	@(#)ufs_disksubr.c	7.15 (Berkeley) %G%
 */

#include "param.h"
#include "systm.h"
#include "buf.h"
#include "disklabel.h"
#include "syslog.h"

/*
 * Seek sort for disks.  We depend on the driver
 * which calls us using b_resid as the current cylinder number.
 *
 * The argument dp structure holds a b_actf activity chain pointer
 * on which we keep two queues, sorted in ascending cylinder order.
 * The first queue holds those requests which are positioned after
 * the current cylinder (in the first request); the second holds
 * requests which came in after their cylinder number was passed.
 * Thus we implement a one way scan, retracting after reaching the
 * end of the drive to the first request on the second queue,
 * at which time it becomes the first queue.
 *
 * A one-way scan is natural because of the way UNIX read-ahead
 * blocks are allocated.
 */

#define	b_cylin	b_resid

disksort(dp, bp)
	register struct buf *dp, *bp;
{
	register struct buf *ap;

	/*
	 * If nothing on the activity queue, then
	 * we become the only thing.
	 */
	ap = dp->b_actf;
	if(ap == NULL) {
		dp->b_actf = bp;
		dp->b_actl = bp;
		bp->av_forw = NULL;
		return;
	}
	/*
	 * If we lie after the first (currently active)
	 * request, then we must locate the second request list
	 * and add ourselves to it.
	 */
	if (bp->b_cylin < ap->b_cylin) {
		while (ap->av_forw) {
			/*
			 * Check for an ``inversion'' in the
			 * normally ascending cylinder numbers,
			 * indicating the start of the second request list.
			 */
			if (ap->av_forw->b_cylin < ap->b_cylin) {
				/*
				 * Search the second request list
				 * for the first request at a larger
				 * cylinder number.  We go before that;
				 * if there is no such request, we go at end.
				 */
				do {
					if (bp->b_cylin < ap->av_forw->b_cylin)
						goto insert;
					if (bp->b_cylin == ap->av_forw->b_cylin &&
					    bp->b_blkno < ap->av_forw->b_blkno)
						goto insert;
					ap = ap->av_forw;
				} while (ap->av_forw);
				goto insert;		/* after last */
			}
			ap = ap->av_forw;
		}
		/*
		 * No inversions... we will go after the last, and
		 * be the first request in the second request list.
		 */
		goto insert;
	}
	/*
	 * Request is at/after the current request...
	 * sort in the first request list.
	 */
	while (ap->av_forw) {
		/*
		 * We want to go after the current request
		 * if there is an inversion after it (i.e. it is
		 * the end of the first request list), or if
		 * the next request is a larger cylinder than our request.
		 */
		if (ap->av_forw->b_cylin < ap->b_cylin ||
		    bp->b_cylin < ap->av_forw->b_cylin ||
		    (bp->b_cylin == ap->av_forw->b_cylin &&
		    bp->b_blkno < ap->av_forw->b_blkno))
			goto insert;
		ap = ap->av_forw;
	}
	/*
	 * Neither a second list nor a larger
	 * request... we go at the end of the first list,
	 * which is the same as the end of the whole schebang.
	 */
insert:
	bp->av_forw = ap->av_forw;
	ap->av_forw = bp;
	if (ap == dp->b_actl)
		dp->b_actl = bp;
}

/*
 * Attempt to read a disk label from a device
 * using the indicated stategy routine.
 * The label must be partly set up before this:
 * secpercyl and anything required in the strategy routine
 * (e.g., sector size) must be filled in before calling us.
 * Returns null on success and an error string on failure.
 */
char *
readdisklabel(dev, strat, lp)
	dev_t dev;
	int (*strat)();
	register struct disklabel *lp;
{
	register struct buf *bp;
	struct disklabel *dlp;
	char *msg = NULL;

	if (lp->d_secperunit == 0)
		lp->d_secperunit = 0x1fffffff;
	lp->d_npartitions = 1;
	if (lp->d_partitions[0].p_size == 0)
		lp->d_partitions[0].p_size = 0x1fffffff;
	lp->d_partitions[0].p_offset = 0;

	bp = geteblk((int)lp->d_secsize);
	bp->b_dev = dev;
	bp->b_blkno = LABELSECTOR;
	bp->b_bcount = lp->d_secsize;
	bp->b_flags = B_BUSY | B_READ;
	bp->b_cylin = LABELSECTOR / lp->d_secpercyl;
	(*strat)(bp);
	if (biowait(bp)) {
		msg = "I/O error";
	} else for (dlp = (struct disklabel *)bp->b_un.b_addr;
	    dlp <= (struct disklabel *)(bp->b_un.b_addr+DEV_BSIZE-sizeof(*dlp));
	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
		if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
			if (msg == NULL)
				msg = "no disk label";
		} else if (dlp->d_npartitions > MAXPARTITIONS ||
			   dkcksum(dlp) != 0)
			msg = "disk label corrupted";
		else {
			*lp = *dlp;
			msg = NULL;
			break;
		}
	}
	bp->b_flags = B_INVAL | B_AGE;
	brelse(bp);
	return (msg);
}

/*
 * Check new disk label for sensibility
 * before setting it.
 */
setdisklabel(olp, nlp, openmask)
	register struct disklabel *olp, *nlp;
	u_long openmask;
{
	register i;
	register struct partition *opp, *npp;

	if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
	    dkcksum(nlp) != 0)
		return (EINVAL);
	while ((i = ffs((long)openmask)) != 0) {
		i--;
		openmask &= ~(1 << i);
		if (nlp->d_npartitions <= i)
			return (EBUSY);
		opp = &olp->d_partitions[i];
		npp = &nlp->d_partitions[i];
		if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
			return (EBUSY);
		/*
		 * Copy internally-set partition information
		 * if new label doesn't include it.		XXX
		 */
		if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
			npp->p_fstype = opp->p_fstype;
			npp->p_fsize = opp->p_fsize;
			npp->p_frag = opp->p_frag;
			npp->p_cpg = opp->p_cpg;
		}
	}
 	nlp->d_checksum = 0;
 	nlp->d_checksum = dkcksum(nlp);
	*olp = *nlp;
	return (0);
}

/* encoding of disk minor numbers, should be elsewhere... */
#define dkunit(dev)		(minor(dev) >> 3)
#define dkpart(dev)		(minor(dev) & 07)
#define dkminor(unit, part)	(((unit) << 3) | (part))

/*
 * Write disk label back to device after modification.
 */
writedisklabel(dev, strat, lp)
	dev_t dev;
	int (*strat)();
	register struct disklabel *lp;
{
	struct buf *bp;
	struct disklabel *dlp;
	int labelpart;
	int error = 0;

	labelpart = dkpart(dev);
	if (lp->d_partitions[labelpart].p_offset != 0) {
		if (lp->d_partitions[0].p_offset != 0)
			return (EXDEV);			/* not quite right */
		labelpart = 0;
	}
	bp = geteblk((int)lp->d_secsize);
	bp->b_dev = makedev(major(dev), dkminor(dkunit(dev), labelpart));
	bp->b_blkno = LABELSECTOR;
	bp->b_bcount = lp->d_secsize;
	bp->b_flags = B_READ;
	(*strat)(bp);
	if (error = biowait(bp))
		goto done;
	for (dlp = (struct disklabel *)bp->b_un.b_addr;
	    dlp <= (struct disklabel *)
	      (bp->b_un.b_addr + lp->d_secsize - sizeof(*dlp));
	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
		if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
		    dkcksum(dlp) == 0) {
			*dlp = *lp;
			bp->b_flags = B_WRITE;
			(*strat)(bp);
			error = biowait(bp);
			goto done;
		}
	}
	error = ESRCH;
done:
	brelse(bp);
	return (error);
}

/*
 * Compute checksum for disk label.
 */
dkcksum(lp)
	register struct disklabel *lp;
{
	register u_short *start, *end;
	register u_short sum = 0;

	start = (u_short *)lp;
	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
	while (start < end)
		sum ^= *start++;
	return (sum);
}

/*
 * Disk error is the preface to plaintive error messages
 * about failing disk transfers.  It prints messages of the form

hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)

 * if the offset of the error in the transfer and a disk label
 * are both available.  blkdone should be -1 if the position of the error
 * is unknown; the disklabel pointer may be null from drivers that have not
 * been converted to use them.  The message is printed with printf
 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
 * The message should be completed (with at least a newline) with printf
 * or addlog, respectively.  There is no trailing space.
 */
diskerr(bp, dname, what, pri, blkdone, lp)
	register struct buf *bp;
	char *dname, *what;
	int pri, blkdone;
	register struct disklabel *lp;
{
	int unit = dkunit(bp->b_dev), part = dkpart(bp->b_dev);
	register int (*pr)(), sn;
	char partname = 'a' + part;
	extern printf(), addlog();

	if (pri != LOG_PRINTF) {
		log(pri, "");
		pr = addlog;
	} else
		pr = printf;
	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
	    bp->b_flags & B_READ ? "read" : "writ");
	sn = bp->b_blkno;
	if (bp->b_bcount <= DEV_BSIZE)
		(*pr)("%d", sn);
	else {
		if (blkdone >= 0) {
			sn += blkdone;
			(*pr)("%d of ", sn);
		}
		(*pr)("%d-%d", bp->b_blkno,
		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
	}
	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
#ifdef tahoe
		sn *= DEV_BSIZE / lp->d_secsize;		/* XXX */
#endif
		sn += lp->d_partitions[part].p_offset;
		(*pr)(" (%s%d bn %d; cn %d", dname, unit, sn,
		    sn / lp->d_secpercyl);
		sn %= lp->d_secpercyl;
		(*pr)(" tn %d sn %d)", sn / lp->d_nsectors, sn % lp->d_nsectors);
	}
}
Commit	Line	Data
da7c5cc6	1	/*
1ad7f5c8	2	* Copyright (c) 1982, 1986, 1988 Regents of the University of California.
7188ac27	3	* All rights reserved.
da7c5cc6	4	*
b702c21d	5	* %sccs.include.redist.c%
7188ac27	6	*
c6f5111d	7	* @(#)ufs_disksubr.c 7.15 (Berkeley) %G%
da7c5cc6	8	*/
200a8f27	9
97934dfa MK	10	#include "param.h"
	11	#include "systm.h"
	12	#include "buf.h"
	13	#include "disklabel.h"
1ad7f5c8	14	#include "syslog.h"
9573c71c	15
200a8f27	16	/*
0d67fde4 BJ	17	* Seek sort for disks. We depend on the driver
	18	* which calls us using b_resid as the current cylinder number.
	19	*
	20	* The argument dp structure holds a b_actf activity chain pointer
	21	* on which we keep two queues, sorted in ascending cylinder order.
	22	* The first queue holds those requests which are positioned after
	23	* the current cylinder (in the first request); the second holds
	24	* requests which came in after their cylinder number was passed.
	25	* Thus we implement a one way scan, retracting after reaching the
	26	* end of the drive to the first request on the second queue,
	27	* at which time it becomes the first queue.
	28	*
	29	* A one-way scan is natural because of the way UNIX read-ahead
	30	* blocks are allocated.
200a8f27 BJ	31	*/
200a8f27 BJ	32
200a8f27 BJ	33	#define b_cylin b_resid
	34
	35	disksort(dp, bp)
0d67fde4	36	register struct buf dp, bp;
200a8f27 BJ	37	{
200a8f27 BJ	38	register struct buf *ap;
200a8f27	39
0d67fde4 BJ	40	/*
	41	* If nothing on the activity queue, then
	42	* we become the only thing.
	43	*/
200a8f27 BJ	44	ap = dp->b_actf;
	45	if(ap == NULL) {
	46	dp->b_actf = bp;
	47	dp->b_actl = bp;
	48	bp->av_forw = NULL;
	49	return;
	50	}
0d67fde4 BJ	51	/*
	52	* If we lie after the first (currently active)
	53	* request, then we must locate the second request list
	54	* and add ourselves to it.
	55	*/
	56	if (bp->b_cylin < ap->b_cylin) {
	57	while (ap->av_forw) {
	58	/*
	59	* Check for an ``inversion'' in the
	60	* normally ascending cylinder numbers,
	61	* indicating the start of the second request list.
	62	*/
	63	if (ap->av_forw->b_cylin < ap->b_cylin) {
	64	/*
	65	* Search the second request list
	66	* for the first request at a larger
	67	* cylinder number. We go before that;
	68	* if there is no such request, we go at end.
	69	*/
	70	do {
	71	if (bp->b_cylin < ap->av_forw->b_cylin)
	72	goto insert;
667e5665 MK	73	if (bp->b_cylin == ap->av_forw->b_cylin &&
	74	bp->b_blkno < ap->av_forw->b_blkno)
	75	goto insert;
0d67fde4 BJ	76	ap = ap->av_forw;
	77	} while (ap->av_forw);
	78	goto insert; /* after last */
	79	}
	80	ap = ap->av_forw;
200a8f27	81	}
0d67fde4 BJ	82	/*
	83	* No inversions... we will go after the last, and
	84	* be the first request in the second request list.
	85	*/
	86	goto insert;
200a8f27	87	}
0d67fde4 BJ	88	/*
	89	* Request is at/after the current request...
	90	* sort in the first request list.
	91	*/
	92	while (ap->av_forw) {
	93	/*
	94	* We want to go after the current request
	95	* if there is an inversion after it (i.e. it is
	96	* the end of the first request list), or if
	97	* the next request is a larger cylinder than our request.
	98	*/
	99	if (ap->av_forw->b_cylin < ap->b_cylin \|\|
667e5665 MK	100	bp->b_cylin < ap->av_forw->b_cylin \|\|
	101	(bp->b_cylin == ap->av_forw->b_cylin &&
	102	bp->b_blkno < ap->av_forw->b_blkno))
0d67fde4 BJ	103	goto insert;
	104	ap = ap->av_forw;
	105	}
	106	/*
	107	* Neither a second list nor a larger
	108	* request... we go at the end of the first list,
	109	* which is the same as the end of the whole schebang.
	110	*/
	111	insert:
	112	bp->av_forw = ap->av_forw;
	113	ap->av_forw = bp;
	114	if (ap == dp->b_actl)
200a8f27 BJ	115	dp->b_actl = bp;
200a8f27 BJ	116	}
97934dfa	117
9573c71c MK	118	/*
	119	* Attempt to read a disk label from a device
	120	* using the indicated stategy routine.
	121	* The label must be partly set up before this:
	122	* secpercyl and anything required in the strategy routine
	123	* (e.g., sector size) must be filled in before calling us.
	124	* Returns null on success and an error string on failure.
	125	*/
	126	char *
	127	readdisklabel(dev, strat, lp)
	128	dev_t dev;
	129	int (*strat)();
	130	register struct disklabel *lp;
	131	{
	132	register struct buf *bp;
	133	struct disklabel *dlp;
	134	char *msg = NULL;
	135
	136	if (lp->d_secperunit == 0)
	137	lp->d_secperunit = 0x1fffffff;
	138	lp->d_npartitions = 1;
	139	if (lp->d_partitions[0].p_size == 0)
	140	lp->d_partitions[0].p_size = 0x1fffffff;
	141	lp->d_partitions[0].p_offset = 0;
	142
d3633d66	143	bp = geteblk((int)lp->d_secsize);
9573c71c MK	144	bp->b_dev = dev;
9573c71c MK	145	bp->b_blkno = LABELSECTOR;
cd6cd535	146	bp->b_bcount = lp->d_secsize;
9573c71c MK	147	bp->b_flags = B_BUSY \| B_READ;
	148	bp->b_cylin = LABELSECTOR / lp->d_secpercyl;
	149	(*strat)(bp);
7188ac27	150	if (biowait(bp)) {
9573c71c	151	msg = "I/O error";
cd6cd535 MK	152	} else for (dlp = (struct disklabel *)bp->b_un.b_addr;
	153	dlp <= (struct disklabel )(bp->b_un.b_addr+DEV_BSIZE-sizeof(dlp));
	154	dlp = (struct disklabel )((char )dlp + sizeof(long))) {
	155	if (dlp->d_magic != DISKMAGIC \|\| dlp->d_magic2 != DISKMAGIC) {
	156	if (msg == NULL)
	157	msg = "no disk label";
0328954d KM	158	} else if (dlp->d_npartitions > MAXPARTITIONS \|\|
0328954d KM	159	dkcksum(dlp) != 0)
9573c71c	160	msg = "disk label corrupted";
cd6cd535	161	else {
9573c71c	162	lp = dlp;
cd6cd535 MK	163	msg = NULL;
	164	break;
	165	}
9573c71c MK	166	}
	167	bp->b_flags = B_INVAL \| B_AGE;
	168	brelse(bp);
	169	return (msg);
	170	}
	171
667e5665 MK	172	/*
	173	* Check new disk label for sensibility
	174	* before setting it.
	175	*/
	176	setdisklabel(olp, nlp, openmask)
	177	register struct disklabel olp, nlp;
	178	u_long openmask;
	179	{
	180	register i;
	181	register struct partition opp, npp;
	182
	183	if (nlp->d_magic != DISKMAGIC \|\| nlp->d_magic2 != DISKMAGIC \|\|
	184	dkcksum(nlp) != 0)
	185	return (EINVAL);
5adcb337	186	while ((i = ffs((long)openmask)) != 0) {
667e5665 MK	187	i--;
	188	openmask &= ~(1 << i);
	189	if (nlp->d_npartitions <= i)
	190	return (EBUSY);
	191	opp = &olp->d_partitions[i];
	192	npp = &nlp->d_partitions[i];
	193	if (npp->p_offset != opp->p_offset \|\| npp->p_size < opp->p_size)
	194	return (EBUSY);
	195	/*
	196	* Copy internally-set partition information
	197	* if new label doesn't include it. XXX
	198	*/
	199	if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
	200	npp->p_fstype = opp->p_fstype;
	201	npp->p_fsize = opp->p_fsize;
	202	npp->p_frag = opp->p_frag;
	203	npp->p_cpg = opp->p_cpg;
	204	}
	205	}
d3633d66 MK	206	nlp->d_checksum = 0;
d3633d66 MK	207	nlp->d_checksum = dkcksum(nlp);
667e5665 MK	208	olp = nlp;
	209	return (0);
	210	}
	211
	212	/* encoding of disk minor numbers, should be elsewhere... */
	213	#define dkunit(dev) (minor(dev) >> 3)
	214	#define dkpart(dev) (minor(dev) & 07)
	215	#define dkminor(unit, part) (((unit) << 3) \| (part))
	216
	217	/*
	218	* Write disk label back to device after modification.
	219	*/
	220	writedisklabel(dev, strat, lp)
	221	dev_t dev;
	222	int (*strat)();
	223	register struct disklabel *lp;
	224	{
	225	struct buf *bp;
	226	struct disklabel *dlp;
	227	int labelpart;
	228	int error = 0;
	229
	230	labelpart = dkpart(dev);
	231	if (lp->d_partitions[labelpart].p_offset != 0) {
	232	if (lp->d_partitions[0].p_offset != 0)
	233	return (EXDEV); /* not quite right */
	234	labelpart = 0;
	235	}
d3633d66	236	bp = geteblk((int)lp->d_secsize);
667e5665 MK	237	bp->b_dev = makedev(major(dev), dkminor(dkunit(dev), labelpart));
	238	bp->b_blkno = LABELSECTOR;
	239	bp->b_bcount = lp->d_secsize;
	240	bp->b_flags = B_READ;
	241	(*strat)(bp);
7188ac27	242	if (error = biowait(bp))
d3633d66	243	goto done;
d3633d66 MK	244	for (dlp = (struct disklabel *)bp->b_un.b_addr;
	245	dlp <= (struct disklabel *)
	246	(bp->b_un.b_addr + lp->d_secsize - sizeof(*dlp));
	247	dlp = (struct disklabel )((char )dlp + sizeof(long))) {
	248	if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
	249	dkcksum(dlp) == 0) {
	250	dlp = lp;
	251	bp->b_flags = B_WRITE;
	252	(*strat)(bp);
7188ac27	253	error = biowait(bp);
d3633d66 MK	254	goto done;
d3633d66 MK	255	}
667e5665	256	}
d3633d66 MK	257	error = ESRCH;
d3633d66 MK	258	done:
667e5665 MK	259	brelse(bp);
	260	return (error);
	261	}
	262
97934dfa MK	263	/*
	264	* Compute checksum for disk label.
	265	*/
	266	dkcksum(lp)
	267	register struct disklabel *lp;
	268	{
	269	register u_short start, end;
	270	register u_short sum = 0;
	271
	272	start = (u_short *)lp;
	273	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
	274	while (start < end)
	275	sum ^= *start++;
	276	return (sum);
	277	}
1ad7f5c8 MK	278
	279	/*
	280	* Disk error is the preface to plaintive error messages
	281	* about failing disk transfers. It prints messages of the form
05de9101 MK	282
	283	hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
	284
1ad7f5c8 MK	285	* if the offset of the error in the transfer and a disk label
	286	* are both available. blkdone should be -1 if the position of the error
	287	* is unknown; the disklabel pointer may be null from drivers that have not
	288	* been converted to use them. The message is printed with printf
	289	* if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
	290	* The message should be completed (with at least a newline) with printf
	291	* or addlog, respectively. There is no trailing space.
	292	*/
	293	diskerr(bp, dname, what, pri, blkdone, lp)
	294	register struct buf *bp;
	295	char dname, what;
	296	int pri, blkdone;
	297	register struct disklabel *lp;
	298	{
	299	int unit = dkunit(bp->b_dev), part = dkpart(bp->b_dev);
	300	register int (*pr)(), sn;
	301	char partname = 'a' + part;
	302	extern printf(), addlog();
	303
	304	if (pri != LOG_PRINTF) {
	305	log(pri, "");
	306	pr = addlog;
	307	} else
	308	pr = printf;
	309	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
	310	bp->b_flags & B_READ ? "read" : "writ");
	311	sn = bp->b_blkno;
	312	if (bp->b_bcount <= DEV_BSIZE)
	313	(*pr)("%d", sn);
	314	else {
	315	if (blkdone >= 0) {
	316	sn += blkdone;
	317	(*pr)("%d of ", sn);
	318	}
	319	(*pr)("%d-%d", bp->b_blkno,
	320	bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
	321	}
f2a9617b MK	322	if (lp && (blkdone >= 0 \|\| bp->b_bcount <= lp->d_secsize)) {
	323	#ifdef tahoe
	324	sn = DEV_BSIZE / lp->d_secsize; / XXX */
	325	#endif
1ad7f5c8	326	sn += lp->d_partitions[part].p_offset;
05de9101 MK	327	(*pr)(" (%s%d bn %d; cn %d", dname, unit, sn,
	328	sn / lp->d_secpercyl);
	329	sn %= lp->d_secpercyl;
9f79ca31	330	(*pr)(" tn %d sn %d)", sn / lp->d_nsectors, sn % lp->d_nsectors);
1ad7f5c8 MK	331	}
1ad7f5c8 MK	332	}